1. Field of the Invention
The present invention relates to a technique of analyzing cause of delay in a single business processing in a job system in which a computer system comprising a plurality of computers connected through a network is used, the computers are respectively assigned to a series of jobs constituting the single business processing, and the computers, to which the jobs in question are assigned, are made to execute respective jobs in a predetermined order, the analysis being carried out associating the cause with processes on the computers and traffic of a network used by the processes.
2. Description of Related Art
As a job system using a computer system comprising a plurality of computers connected through a network, there is known one in which, taking resources and a load of each computer into consideration, jobs are dynamically assigned to respective computers to be executed by them, so that delay in processing is not generated, as described in Japanese Unexamined Patent Laid-Open Nos. 10-11406, 10-161987, and 9-311795.
Further, as a job system using a computer system comprising a plurality of computers connected through a network, there is known one in which, a computer is assigned to each of a series of jobs constituting one business processing, so that the jobs are executed in predetermined order by respective computers to which those jobs are assigned.
On the other hand, as a technique of analyzing a location of a bottleneck to performance, in a computer system comprising a plurality of computers connected through a network, there is known a technique in which pieces of load information about component devices constituting the computer system are collected, and a bottleneck component device is judged from the collected information, as described in Japanese Unexamined Patent Laid-Open No. 8-137810.
Now when cause of delay in a single business processing is to be detected in the above-mentioned job system in which computers are respectively assigned to a series of jobs constituting so that the single business processing and the jobs are executed in predetermined order by respective computers to which those jobs are assigned, it is insufficient to judge a component device that is a bottleneck of performance of the computer system, applying the technique of the above-mentioned Japanese Unexamined Patent Laid-Open No. 8-137810. It is because delay in a business processing in such a job system is generated from relation between component devices of the computer system and jobs. Thus its cause should be detected taking account of the relation between the component devices of the computer system and the jobs, more closely, to be able to decide effective measures to get rid of generated delay.
Further, the job system described in Japanese Unexamined Patent Laid-Open Nos. 10-11406, 10-161987, and 9-311795 is one in which taking resources and a load of each computer into consideration, jobs are dynamically assigned to respective computers to be executed by them, so that delay in processing is not generated. Thus, in a job system in which a system designer assigns computers to a series of jobs and makes the jobs be executed in a predetermined order by respective computers to which the jobs in question are assigned, it is not supposed to present necessary material for reconsideration when the system designer reconsiders assignment of jobs to respective computers.
Accordingly, an object of the present invention is to effectively detect a causative location, in relation to loads of the computers, network traffic, and the jobs.
Further, another object of the present invention is to present necessary information for reconsideration when the assignment of jobs to respective computers constituting the above-mentioned job system is reconsidered.
To attain the above objects, the present invention provides a method of analyzing a delay factor, for analyzing a delay factor in processing a job net in a job system that uses a computer system having a plurality of computers connected, for example, through a network, assigns the computers respectively to a series of jobs constituting the job net, and executes the jobs in a predetermined order by the respective computers assigned to the jobs, to process the job net, wherein the method comprises:
a collecting step in which history information expressing history of execution of a job from each computer is collected; and
an analyzing step in which a job and a part of the computer system that became delay factors in processing the job net is decided, in accordance with a form of delay in execution of each job expressed in the history, in relation to an execution schedule of each job expressed in definition information defined in advance, and in accordance with computers assigned to each job.
Here, a job net implies a unified processing (for example, business processing) that is constituted by a series of jobs and processed by the job system.
According to thus-described method, it is possible to extract a part of the computers, which became a delay factor in processing the job net, in relation to the jobs.
In detail, the analyzing step comprises:
a first step in which a computer assigned to a job whose execution time exceeds a planned execution time by more than a predetermined degree is extracted as a delay factor in processing the job net; and
a second step in which, with respect to a job whose execution end time is latest among jobs executed just prior to a job whose execution start time is delayed from a planned start time by more than a predetermined degree, when the execution end time is not delayed from a planned end time by more than a predetermined degree, the part of the computer system that undertakes transfer between the job whose execution start time is delayed from the planned start time by more than the predetermined degree and the job executed just prior to the job in question is extracted as a delay factor in processing the job net.
By doing this, it is possible to present a job that became a delay factor in processing a job net, and at the same time, to distinguish a case where that factor was caused by a computer, and a case where a part of the above-mentioned computer system that undertakes transfer between two jobs.
Further, to attain the above objects, the present invention provides a method of analyzing a delay factor, for analyzing a delay factor in processing a job net in a job system that uses a computer system having a plurality of computers connected through a network, assigns the computers respectively to a series of jobs constituting the job net, and executes the jobs in a predetermined order by the respective computers assigned to the jobs, to process the job net, wherein the method comprises:
a collecting step in which history information from each computer is collected, the history information being information that can specify an execution start time at which the computer started execution of a job, and an execution time which the computer required to execute the job, from each computer; and
an analyzing step in which the delay factor in processing the job net is decided by comparing definition information, which is defined in advance and can specify a planned start time, i.e., a planned time at which execution of each job is started, and a planned execution time, i.e., a planned time required for execution of each job, with the history information collected, and
the analyzing step comprises:
a first step, in which a specific job is noticed in accordance with a predetermined rule;
a second step, for repeating a process in which, when an execution start time of the noticed job is delayed from a planned start time of the job by more than a predetermined degree, a job executed just prior to the noticed job in the order is noticed; and
a third step, in which, with regard to each noticed job, when an execution time of the job exceeds a planned execution time of the job by a predetermined degree, the job is extracted as the delay factor in processing the job net.
According to thus-described method, when an execution start time of the noticed job does not exceed the planned start time by more than the predetermined degree, it is possible to exclude jobs earlier than the noticed job in that order, from the processing of searching a delay factor, judging that the delay factor does not exist in those jobs. Thus, a delay factor can be searched more efficiently.
Further, to attain the above objects, the present invention provides a method of analyzing a delay factor, for analyzing a delay factor in processing a job net in a job system that uses a computer system having a plurality of computers connected through a network, assigns the computers respectively to a series of jobs constituting the job net, and executes the jobs in a predetermined order by the respective computers assigned to the jobs, to process the job net, wherein the method comprises:
a collecting step in which history information is collected from each computer, the history information being information that can specify an execution start time at which the computer started execution of a job, an execution end time at which the computer ended the execution of the job, and an execution time which the computer required to execute the job, from each computer; and
an analyzing step in which the delay factor in processing the job net is decided by comparing definition information, which is defined in advance and can specify a planned start time, i.e., a planned time at which execution of each job is started, a planned end time, i.e., a planned time at which execution of each job is ended, and a planned execution time, i.e., a planned time required for execution of each job, with the history information collected, and the analyzing step comprises:
a first step, in which a specific job is noticed in accordance with a predetermined rule;
a second step, for repeating a process in which, among jobs executed just prior to the noticed job in the order, jobs whose execution end times stand within n-th (n: a natural number) from the latest are extracted and noticed; and
a third step, in which, with regard to each noticed job, when an execution time of the job exceeds a planned execution time of the job by a predetermined degree, the job is extracted as the delay factor in processing the job net.
According to thus-described method, it is judged that, among the jobs executed just prior to the noticed job in that order, jobs having later execution end times become bottlenecks affecting the delay as a whole. Among the jobs executed just prior to the noticed job, jobs having earlier execution end times are judged not to be a delay factor, and can be excluded from the processing of searching a delay factor. Thus, it is possible to efficiently search a delay factor.
Further, to attain the above objects, the present invention provides a method of analyzing a delay factor, for analyzing a delay factor in processing a job net in a job system that uses a computer system having a plurality of computers connected through a network, assigns the computers respectively to a series of jobs constituting the job net, and executes the jobs in a predetermined order by the respective computers assigned to the jobs, to process the job net, wherein the method comprises:
a collecting step in which history information is collected from each computer, the history information expressing a history of executing a job which constitutes the job net and is assigned to the computer;
an analyzing step in which a job and a computer assigned to the job that became the delay factors in processing the job net are decided, in accordance with a form of delay in execution of each of the series of jobs constituting the job net, expressed by the history information collected, in relation to an execution schedule of each of the series of jobs constituting the job net, expressed in definition information defined in advance, and in accordance with computers assigned to each of the series of jobs;
a computer utilization status acquiring step in which a resource utilization status owing to each process executed in the computer to which the job became the delay factor in processing the job net is assigned is acquired, from the computer;
a process specifying step in which a process as an execution form of the job that became the delay factor in processing the job net is specified, on the computer to which the job is assigned, and a computer utilization status of the computer owing to the process of the job that became the delay factor in processing the job net is specified, among resource utilization statuses owing to respective processes acquired from the computer to which the job that became the delay factor in processing the job net is assigned; and
a computer utilization status displaying step in which a resource utilization status of the computer to which the job that became the delay factor in processing the job net is assigned is displayed, owing to each process, acquired from the computer, distinguishing, at least, the resource utilization status of the computer owing to the process specified with respect to the job that became the delay factor in processing the job net.
According to thus-described method, since a resource utilization status of a computer to which a job detected as a factor of delay in processing a job net is assigned is displayed, distinguishing, at least, the resource utilization status owing to a process of the job, it is easy to understand if the job presses a load of the computer, or another job or application program presses the load of the computer. In other words, it is possible to present necessary information when assignment of jobs to respective computers constituting a job system is to be reconsidered.
Further, to attain the above objects, the present invention provides a method of analyzing a delay factor, for analyzing a delay factor in processing a job net in a job system that uses a computer system having a plurality of computers connected through a network, assigns the computers respectively to a series of jobs constituting the job net, and executes the jobs in a predetermined order by the respective computers assigned to the jobs, to process the job net, wherein the method comprises:
a collecting step in which history information is collected from each computer, the history information expressing a history of executing a job which constitutes the job net and is assigned to the computer; and
an analyzing step in which a job and a computer assigned to the job that became the delay factors in processing the job net are decided, in accordance with a form of delay in execution of each of the series of jobs constituting the job net, expressed by the history information collected, in relation to an execution schedule of each of the series of jobs constituting the job net, expressed in definition information defined in advance, and in accordance with computers assigned to each of the series of jobs;
a network specifying step in which a process as an execution form of the job that became the delay factor in processing the job net is specified, on the computer to which the job is assigned, and a logical channel used by the specified process and a part of the network to which the logical channel is set are specified;
a traffic status collecting step in which a traffic status in the specified network is collected, for each logical channel set to the network; and
a traffic status displaying step in which the traffic status collected for the each logical channel is displayed, so as to, at least, distinguish a traffic status of the logical channel used by the process specified to the job that became the delay factor in processing the job net.
Here, the logical channel used by the specified process implies a logical channel on the network, which is set between the process and an application, when the job corresponding to the process requests processing by that application running on a computer different from a job executing computer that executes the process.
According to thus-described method, a traffic status of a network to which a logical channel used by a process of a job detected as a delay factor in processing a job net is set is displayed, distinguishing, as least, a traffic status of a logical channel used by the process of the job. Thus, it is easy to understand if the job presses a load of the network, or another job or application presses the load of the network. In other words, it is possible to present necessary information when assignment of jobs to respective computers constituting a job system is to be reconsidered.